Can a uni reliably say someone has used ai to create their work?

[Unissss]

i personally don’t see how tbh

I work at a university and yes they can mainly tell. Sure some slip through but turnitin is pretty good but one academic was telling me students often leave in the prompts or the 'ask me more' like. Last year she had four students hand in the exact same essay. And often the bibliography is made up too and the papers don't exist

The short answer is yes, a lecturer can fairly easily tell if you've used AI. But there's generally not a lot they can do about it if the student really digs in and brings a lawyer along to the disciplinary. My daughter's university students have learned to do that, and the university always has to back down as they can't afford lawsuits.

I write books for a living. If I find myself accused of using AI, as authors increasingly are, whether they've used it or not, I could show my entire working process with one hundred odd drafts. That will soon not be enough as people are learning to generate draft Word documents showing fake developmental processes, using AI.

But what I could easily do is sit down for an hour under controlled conditions and write a chapter about my characters facing any situation the examiner cares to come up with. The style and humour and use of language would match all my previous works perfectly.

That's the kind of thing that universities are going to have to work out how to do in the future. Not just exams and timed essays under controlled conditions, but rewriting a portion of an essay and replicating the arguments and structure.

AI essays are very similar to each other, especially in structure and citations (which are very often made up). AI also has a writing ‘style’ which is easy to notice - once you see it you can’t unsee it. We’re using vivas more and more to make sure the student understands what they’ve written.

Uni lecturer here. I get this is an internet forum, but your spelling/grammar leave a lot to be desired. Why not just use the word spelling and grammar checker? Running your essay through AI is an academic offense, and if they have pulled you on it you can't have any complaints.

Nope. They'll keep taking your 9 grand, you'll leave with an average grade and I wouldn't go to them for a reference when you leave. Just this week, a colleague has flagged a student's AI use to a potential employer.

Yes. There are numerous red flags.

If you're asking, "Can they point to one single thing in the work which irrefutably and beyond all reasonable doubt indicates that AI was used?" then no, that's not possible.

But the red flags add up.

The clincher comes when the student cannot explain what they wrote, or cannot replicate the same quality of writing when writing by hand under supervision.

My siblings works in a university leading / coordinating this specific area.

Yes.

I'm bored with the minutae of it frequently and also many many students arent that sophisticated in their efforts.

The most most basic conversation thats has been had eleventh billion times

Sibling: "Hi x, can you please type a double dash on your keyboard here infront of me?"

Student: <looks blankly feeble attempts it and fails a few times>

Sibling: "it's surprising you cannot type a double dash youve used it 42 times in this essay. You've written eloquently about X can you tell me <basic question>"

Student: <blank expression and silence... occasionally reads verbatim something from the chatgpt essay they are holding>
Sibling: 🫠🫠🫠🫠

The rampant use of AI for cheating inevitably means some innocent people will be suspected. This is not the fault of the academics who are fighting a rearguard action to ensure that the degrees people are paying so much for retain their value. Any individual who really knows their stuff and can defend it in a viva voce will be just fine.

Lazy, lying, cheating dimwits..

Uni is FOR reading, critical thinking, new ideas, further research, applying acquired knowledge, exploring concepts...

What is this fresh dystopian hell and what the fuck are they DOING with their time?!

We had basic google and even that was shit in the 90s. At least people will always respect my degree, post grad, Diploma and Masters for what it is. Hard fucking work and ability.

AI is hard to detect in student essays because it can write in a very natural, human-like way without leaving clear clues behind. The content can also be tweaked or rewritten so it doesn’t look repetitive, and a lot of AI-generated work ends up sounding like something a good student would produce anyway. On top of that, detection tools aren’t fully reliable, so it’s tough to confidently tell the difference between AI writing and genuine student work.

Ironically I think there are ai computer programmes that check for this. Almost like marking their own homework. As ai gets better I think it will become harder. I imagine it’s like the plagiarism software from 20 years ago.

However, I work in an area where I deal with complaints and grievances etc. I can usually tell when I believe something is written by ai. Often it is the structure eg a heading followed by four bulletpoints, another heading and four bulletpoints etc. I also sometimes think it when the document quotes unnecessary legal jargon and doesn’t get to the point quickly. I’ve gone from receiving complaints that are 4 pages long and clearly state the desired resolution to 20 plus pages of pure waffle in the last 12 months.

I find asking them to type an em-dash works pretty well too. (I’m a teacher in secondary school.)

Not a single kid I have suspected of using AI has been able to type an em-dash, despite their work being full of them. (It’s the long hyphen thingy; there’s no button for it on the keyboard and typing one is very convoluted.)

Is it really? Im a doctoral student (UCL) and yesterday i was discussing with my research supervisor and another student how to use AI for proof reading, word limit and structuring of dissertation….
as long as it is not use to generate ideas, it is ok to use
also only using the UCL co-pilot Ai but not any other LLMs dur to confidentiality issues

Just admit to your uni that you used AI for spelling and grammar. Get the version you put into AI and instead correct via MS Word spell and grammar checker. Then hand back in.

Nothing that will stand up in court

I had a meeting with an AI checker software sales team recently and they were very clear about that.

There’s just a ‘feel’ to it, and once you’re wise to it - you can’t unsee it.

I couldn’t agree more.

People in fairly senior positions at work were using it to write report.
Which i took issue with, cos if you’re on £60k a year, due to your skills and seniority within the company/industry, then one of the tasks that shouldn’t be beyond your wit, is writing a report.

Lots of people disagreed with me, for spurious reasons. They didn’t change my mind.

I was in a café the other day and saw a student next to me sit with two windows open on his laptop. On one he prompted Chatgpt to write a section for his essay and in the other he was essentially rewriting the chatgpt output in his own words. Yes he was rewriting the text but he wasn't using his own notes and so did he do any of his own reading or research? Its hard to say. I also noticed that he stopped every few lines to check his phone and scroll though Instagram and other socials before going back to his essay then after another line he'd be back to his phone and so on. Yeah it was pretty nosy of me but it was interesting to see how he was interacting with the technology. I don't know if his use is typical of students or if rewriting really makes a difference to comprehension or detection? Can AI do all that and provide proper footnotes and a bibliography? Is all that still required?

A friend of mine works at a University and says its a huge issue and there seems to be a lot of debate on how to handle it with some being very permissive i.e. its the future embrace it and others being quite hostile and pointing to studies which suggest it is literally making students dumber. As a former student I can see the temptation for sure.

Recently I put a uni question into AI just to see what it would say, but I wrote my own answer, we had to share our answers on a group board with other students and someone put the exact AI answer I had got, with random words in bold too!

I don't think they could conclusively. If a student is clever and uses it from the very beginning it is hard to prove it isn't just their writing style (comparing exam writing to coursework writing is't a fair test either). AI didn't exist when I was at uni 25 years ago but the essays I submitted as coursework would be far more articulate than what I could write in an exam or explain verbally, there was no cheating going on in my case, it wasn't possible back then. I am dyslexic though, so having the luxury of time when writing my coursework meant I could spend ages rewriting it so it flowed much better than my exam style writing (I was only tested for dyslexia during my PhD so didn't have extra exam time for undergrad or masters).

You don't need AI for a spelling and grammar check. Just use Grammarly. It's free.

It's tricky because I use AI for work quite often to refine what I want to say when I have some writer's block. So I will type in what I want to say even if I know it's not worded well and it normally refines it.

I always make changes to what is produced but I wouldn't copy over any text to an essay for uni because of the risk of copying code.

There is a "way" some AI tools write though which introduce Americanisms not used widely in British English. A few times I've noticed what it's come up with is "off" in tone but without being able to put my finger on why. Then there's the ubiquitous "em dash" and words like "quietly" which it seems to love.

Definitively prove it? I have no idea. But especially as it's new, I'm sure lecturers can see sudden changes in style and be suspicious right now. But I think Linkedin is showing the way it might go. If you are on Linkedin much you will see how ubiquitous AI posts are. Even if you can't pin exactly why, you know. And the common thread is? They're really boring. What is now starting to stand out (at least on my feed) is the slightly messy, but authentic. It does make me wonder whether an AI produced essay might get you a 2(ii) or 2(i) grade in an essay but no higher.

Lastly and this is the big kicker for uni work, AI searches for references? Just don't go there. They are sometimes plain wrong, don't exist or don't reference what it claims they do. Or it's a really minor study which is so niche and against the prevailing research. If you get AI to write it all then that's what you'll get and it will be incredibly obvious.

Where I think you could use AI and get away with it is to ask how to structure your work, what sections to include etc and then, without copying that, use that as a basis.

You could just use the inbuilt tools in Word...

Google docs has a function which detects ai

A friend of DC was distraught when she was accused of using AI by her university. She lost marks. Sometimes they get it wrong. This girl most definitely would not and did not use AI.